Discrimination of atrial arrhythmias using autoregressive modelling

Classification of atrial arrhythmias from bipolar endocardial leads is inhibited by distortions in organized rhythms and lack of discrete deflections in atrial fibrillation. Autoregressive modeling of organized rhythms produces impulse-dominated residuals, while atrial fibrillation shows a residual consisting of decorrelated noise. These distinct differences in the residual amplitude distributions permit separation of atrial fibrillation from other atrial arrhythmias. A real-time median adaptive filter produced comparable residuals for the organized rhythms with faster, smoother adaptation than standard least mean square algorithms due to impulse-enhancement of the residuals